Académique Documents
Professionnel Documents
Culture Documents
Article information:
To cite this document:
Patrick Pujo Ilham El Khabous Fouzia Ounnar , (2015),"Experimental assessment of the productivity
improvement when using U-shaped production cells with variable takt time", International Journal of
Lean Six Sigma, Vol. 6 Iss 1 pp. 17 - 38
Permanent link to this document:
http://dx.doi.org/10.1108/IJLSS-07-2013-0038
Downloaded on: 06 February 2016, At: 23:08 (PT)
References: this document contains references to 48 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 134 times since 2015*
Access to this document was granted through an Emerald subscription provided by emeraldsrm:272736 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/2040-4166.htm
Ilham El Khabous
Variable takt
time
17
Received 11 July 2013
Revised 1 May 2014
19 May 2014
Accepted 31 May 2014
Fouzia Ounnar
CRET-LOG, Aix Marseille University, Marseilles, France
Abstract
Purpose The aim of this research is to discuss the benefits of U-shaped layout for production cell
operating in variable takt time. Different experiments were conducted using benchmarks to highlight
the performance gap between a linear cell and a U-Cell.
Design/methodology/approach The implementation of the production cell, either in a U-shaped
or in a straight line layout, is optimized through linear programming based on the number of operators.
The two corresponding programs, in Mosel language, use the same approach to not introduce bias in the
comparison of results. The study used the authors own datasets and other well-known academic
benchmarks.
Findings A comparison was conducted between the obtained takt times, with equivalent operating
conditions, in both U-Cell and linear cell. A significant increase of the production rate was observed.
This increase has often exceeded 10 per cent, reaching 32 per cent. All the experiments show that, with
the same number of operators, a cell in a U-shaped layout is always at least as efficient, in terms of
attainable production rates, than an equivalent linear cell. Ninety-six per cent of the studied cases give
an improvement of production rate. Moreover, the dispersion of the U-Cell results is weaker, which
suggests that the U-shaped layout gives better performances in more robust manner.
Research limitations/implications Results were obtained through a study of various academic
benchmarks. The results must be validated on industrial situations.
Practical implications This paper will be very useful for researchers and practitioners to
understand lean implementations and their derived benefits. This paper will allow them to evaluate and
analyze the expected benefits of the implementation of the production cell in the U-shaped layout
(operating in variable takt time).
Originality/value U-Cells constitute an appropriate solution for a layout of any kind of production
cells with a variable structure (variability of the number of operators, of the organization of the cell, of
The authors thank FICO, which, through its Academic Partner Program, has kindly put the
FICOTM Xpress Optimization Suite at our disposal.
IJLSS
6,1
the takt time []). When facing a significant variation in the demand, the response consists of adjusting
the number of operators assigned to the cell. This study jointly addresses the problem of the U-shaped
layout and the operation in variable takt time.
Keywords Line balancing, Linear programming for optimization, Production variable throughput,
Straight production line, U-shaped production cell
Paper type Research paper
18
Downloaded by FLINDERS UNIVERSITY OF SOUTH AUSTRALIA At 23:08 06 February 2016 (PT)
1. Introduction
Recent developments toward lean management (Holweg, 2007) reflect an increasing
trend by companies to use this approach. This is often found in the literature in the form
of testimonials: the effectiveness of lean management methods and tools is typically
expressed through case studies (Detty and Yingling, 2000; Motwani, 2003; Modarress
et al., 2005; Kumar et al., 2006). For example, tools such as value stream mapping (VSM)
have been the subject of evaluation before and after comparison, used alone (Singh
et al., 2010) or in combination with other techniques in the case of production lines
(Alvarez et al., 2009). Based on VSM and milkrun, using lean metrics, such as
dock-to-dock time and lean rate, these works analyze the internal materials flow and its
improvement.
Other tools, such as U-shaped cells producing or takt time variable, have not been
subjected to comparison before and after.
U-shaped production cells (Figure1(b)) emerged in the Western industrialized world
with the introduction of the Just-in-Time philosophy (Monden, 1983) and are currently
associated with the Lean Production approach (Shah and Ward, 2003).
The majority of the production lines not identified as falling within one of these two
approaches are organized linearly (Figure 1(a)). The optimization of their performance is
obtained using load balancing techniques which are dealt with in numerous studies
(Becker and Scholl, 2006).
U-Cells offer many advantages over traditional linear arrangements, in terms of
flexibility, standardization, decrease in work in progress and the maintaining of
high-quality products. They also provide improvements in productivity.
In the case of high variations in production volumes, it is easy to adjust the number
of operators to adapt the production rate to the instantaneous demand (takt time).
Through variable takt time, the production lines enable new strategies, so that lean
(a)
Figure 1.
Straight production
cell (a), U-shaped
production cell (b)
(b)
product realization can become a reality for variable productions (Weston and Cui,
2008). This is common in U-Cells, but this can also be achieved in straight production
lines. It then becomes interesting to compare both layout approaches and evaluate their
respective performances through experiments using benchmarks.
After defining the basic principles of the operation of a variable rate production line,
a few fundamentals regarding U-Cells are recalled. Then, the issue of the layout
optimization for such systems is addressed, and linear programming models are
developed aiming to obtain the best possible balancing of the operator tasks.
These models are then implemented into a solver and tested on academic
benchmarks which are set to be the most impartial and the least specific possible.
Finally, the results of these experiments are analyzed and discussed.
2. Choice of a layout architecture for a production cell
2.1 Cellular production: some reminders
Production lines were deployed at the beginning of the twentieth century, after Henry
Ford installed the first assembly line in 1913 (Sward, 1989). After being used for a
century in mass manufacturing, this concept is still evolving today (Hirst and Zeitlin,
1991), being revisited (Bartholdi and Eisenstein, 1996). This history, combined with a
pragmatic quest for efficiency (Magee, 2007), led the Japanese industry to develop the
idea of production cells: a production cell is a small production unit which specializes in
the manufacturing of a family of products or parts.
This type of organization should offer greater usage flexibility; better performances
in terms of quality, cost and time; and fluidization of flows combined with reduction in
the volume of work in progress and greater responsiveness.
In fact, a cell is a small production line whose conditions of use can vary according to
needs: change of rate, of products, of the number of machines, etc. The line architecture
is related either to the grouping together of all the processes related to the fabrication of
a family of products or to the grouping resulting from a frequent series of operations
common to the manufacturing of different types of products.
When this approach is generalized to the whole production system, we can refer to it
as cellular manufacturing (Irani, 1999): the variety of products to be manufactured, their
estimated volumes and the fluctuations in demand will determine the set of cells which
will need to be formed and designed. Numerous studies about cellular production focus
on the characterization of this system using different approaches such as group
technology (Kulak et al., 2005), integer linear programming (Ariafar and Ismail, 2009),
metaheuristics of genetic algorithm type (Yin and Yasuda, 2006) or ant colony
optimization algorithms (Sabuncuoglu et al., 2009).
Another emerging problem is related to the dynamics of the control of this set of cells
(Wenming and Deuse, 2009). For instance, if we consider a workshop where many
neighboring cells respond to varying and desynchronized customer loads, it is possible
to adjust the number of operators per cell and transfer operators within cells
(communicating vessels) so that the workshop is fully controlled and maximum
productivity is achieved, preserving responsiveness and dynamics, whatever the
variation of customer demands, and providing a high potential increase in production
rate. For example, a cell functioning with n operators (nominal operation) can be
adjusted to n 1 or n 1, but it is possible to make it function even faster, with up to
Variable takt
time
19
IJLSS
6,1
20
n k operators, the additional operators being taken from neighboring cells or even
make it function very slowly (with only one operator) or stop it temporarily.
In any case, the search for the optimal cell productivity requires the balancing of the
production loads: the key is to set up the cell so it can produce at the lowest cost possible
while complying with the technical constraints.
2.2 Balancing of a production cell
This issue was first raised in automotive assembly lines, and is now referred to, in the
scientific literature, as Simple Assembly Line Balancing Problem (SALBP) (Baybars,
1986). A typology of the resolution methods for this type of problems was suggested by
Boysen et al. (2007). Generally speaking, these methods consist in splitting the task as
effectively as possible between the operators so that the workload is equally distributed
and the production rate is optimal.
The general features of this problem are the following:
a single operation cannot be shared between different operators;
the operations are partially ordered;
all operations must be executed;
durations of operations must not depend either on the operator or on previous
operations;
an operator can carry out any operation;
any operation can be executed by any operator;
the products pass one by one in front of operators in a given order;
the operators themselves are positioned in a given order; and
one single product is produced per line.
Operations have to be split between operators so that waiting times are minimal. The
waiting time of each operator is determined by the difference between the cycle time of
the line (the load of the busiest operator) and the actual load (the sum of all the times of
operation assumed by the operator). The solution in which the operator waiting time is
the shortest provides the lowest cost.
Two cases are possible. First, if the operators are assigned to more or less versatile
workstations (for example, in the case of an assembly workshop), the balancing will
consist in assigning them maximum number of operations. Then, if the operators have
to move between different workstations, the balancing will consist in assigning them
various consecutive workstations, with a displacement cycle between these
workstations.
Even though the mathematical formulation of these two cases is equivalent, our
interest will be mainly focused on the second case. The displacement of operators
between different workstations is called Chaku-Chaku (Wenming and Deuse, 2009).
Let us note that the so-called caravan system (Spengler et al., 2005), which consists
in moving k operators (out of phase by 2/k) along the entire cell (Figure 2(a)), with an
identical Chaku-Chaku, is not taken into account. We will focus on the balancing of the
line between different Chaku-Chaku (one operator corresponding to one Chaku-Chaku),
involving a manual transfer of parts between neighboring Chaku-Chaku operators
(Figure 2(b)).
It is also important to detail the nature of tasks Ti which refer to the operations
carried out by operators Ok on the different workstations Pi. We should distinguish
between durations Doi of tasks Ti and displacement times between workstations Ddxi.
The latter are usually small compared to the cumulative durations of tasks performed
by each workstation, and they can be either neglected or integrated to the corresponding
durations: Doi includes Ddxi. Besides the displacement time, the duration includes
product loading/unloading tasks on the workstation, possible product control tasks and
either the execution of this task in the case of a purely manual task or the initiation
of the execution of the task followed by a short period of monitoring to visually check
whether the task has started correctly in the case of an automated workstation.
This refers to the apparent time component Di of the operation on workstation Pi. For
automated tasks, the continuation and completion of the task progress in masked time,
on a period called Dmi. Note that Doi Di Dmi. To each workstation Pi corresponds
a chronogram of all the tasks of the operation (Monden, 1983), called Standard
Operation Chart, and all the chronograms show the cyclic activity of operator Ok in a
Gantt diagram, called Standard Operation Combination Chart.
If Tk is the cycle time of operator k, the following relationships must be fulfilled:
For a given operator k, to which nk operations i have been assigned:
Variable takt
time
21
nk
D
j1
ij
Tk
(1)
(2)
If T is the cycle time of the cell and np the number of operators, then:
k 1, np , max( Tk ) T
(3)
These two overall variables strongly influence the resolution method, depending on
whether they are fixed or to be minimized. SALBP can be expressed in four ways
(Baybars, 1986; Becker and Scholl, 2006) (Figure 3):
(1) SALBP-F is the feasibility study of a configuration, and not a search for
optimization.
(a)
(b)
Figure 2.
Chaku-Chaku in
caravan system (a),
Chaku-Chaku with
manual transfer (b)
IJLSS
6,1
(2)
(3)
(4)
22
These SALBP problems deal with the balancing of straight cells. The layout of
U-shaped production cells is the subject of specific studies collectively known as Simple
U-Line Balancing Problem (SULBP) (Aase et al., 2004) or U-Assembly Line Balancing
Problem (UALBP) (Scholl and Klein, 1999b).
2.3 Balancing of a U-Cell
For U-shaped cells, the classification of the resolution methods is similar (Kriengkorakot
and Pianthong, 2007; Boysen et al., 2007) identifies a number of jobs, denoted by [|u|].
Regarding SULBP-1, Miltenburg (1998) first proposed a dynamic programming
approach, followed by an integer linear programming approach (Miltenburg, 2001b).
More recently, Sabuncuoglu et al. (2009) suggested a resolution using an ant colony
metaheuristic.
Other studies relate to SULBP-E, for example, Chiang and Urban (2006) used a
heuristic procedure, Kara et al. (2009) applied a multi-objective fuzzy integer
programming method.
Finally, a number of studies address the Mixed Model U-Line Balancing and
Scheduling Problem (MMULBS) regarding the production of a variety of products in a
U-Cell, simultaneously considering balancing and scheduling (Miltenburg, 2002; Kim
et al., 2006; Kazemi et al., 2011).
2.4 Performance comparison between the balancing of a straight cell and a U-shaped
cell
Numerous studies were undertaken simultaneously on both approaches (Agpak and
Gken, 2007, Kara et al., 2009), but focusing on the way the production cells were
organized and thus on the cell architecture generation algorithm, without comparing
their performances with equivalent datasets.
As our study is performed from a perspective of variable takt time, we can compare
both types of layout by changing the number of operators and comparing the respective
production rates obtained. For that, we should move towards SALBP-2 and SULBP-2
approaches.
3. Modeling of the SALBP-2 balancing problem for a production cell
In this section, the implemented linear programming models are described. The two
models have to be homogeneous in terms of structure and design to avoid any
Cycle time (rate)
Figure 3.
Classification of line
balancing problems
Number of workstations
Fixed
Fixed
SALBP-F
To minimize
SALBP-2
To minimize
SALBP-1
SALBP-E
i 1nt :
ik
(4)
k1
np
( u, v ) 1nt :
np
k.x
uk
k1
k.x
vk
(5)
k1
nt
k 1np :
D .x
i
ik
(6)
i1
Min z T
(7)
(8)
T0
(9)
Under these conditions, the first operator handles the cell inputs, while operator np
manages the outputs.
Variable takt
time
23
IJLSS
6,1
24
i 1nt :
ikj
(10)
j1 k1
np
( u, v ) 1nt :
np
k.x
uk1
k1
1nt :
k 1np :
vk2
k1
nt
j1
i1
(11)
k.x
uk2
(12)
(13)
np
k.x
2
vk1
k1
np
( u, v )
k.x
k1
D .x
i
ikj
Min z T
(14)
(15)
T0
(16)
In this formulation of the SULBP-2 problem, operator np manages the inputs and
outputs of the cell. Operator 1 is situated in the heart of the cell.
3.3 Implementation of the models
The implementation of these models is performed using Mosel language and Xpress
solver, available under FICO optimization environment (www.fico.com).
Thereafter, only the programming of the SULBP-2 problem (Figure 4) will be
described, that of the SALBP-2 problem having already been described by Guret et al.
(2002).
The binary assignment variable x is represented by a three-dimensional table (nt, np,
2), the third dimension being used to separately treat the two branches of the U.
Variable takt
time
25
Figure 4.
SULBP-2 problem
programmed in
Mosel language
IJLSS
6,1
26
Table I.
Description of the
dataset: the 34
studied cases
Case
no.
P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
P13
P14
P15
P16
P17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
5
10
10
5
5
5
11
18
17
5
10
14
5
60
20
10
5
10
60
6
5
30
5
8
12
9
15
12
5
7
60
11
6
25
23
9
60
8
10
15
5
60
21
9
5
21
60
8
11
8
20
35
5
5
23
37
12
17
6
6
25
15
20
9
5
6
17
60
12
17
15
5
20
20
12
22
25
16
60
22
51
14
8
15
20
5
40
25
14
22
29
60
15
12
5
5
12
31
11
20
60
10
10
60
5
11
18
60
34
15
5
15
7
60
18
36
5
22
5
60
30
10
25
9
35
6
17
19
32
30
9
21
52
5
5
5
15
5
13
10
25
35
24
9
22
13
16
5
39
13
60
18
5
23
20
15
11
34
8
12
10
5
12
23
16
9
12
26
5
13
5
14
22
12
20
10
18
6
27
60
24
19
6
18
14
5
30
8
10
7
9
11
7
15
60
7
15
8
5
11
32
18
51
16
30
22
30
12
52
8
14
56
14
16
11
27
12
7
30
60
60
20
5
30
7
12
10
9
30
31
11
10
27
10
5
30
9
33
8
50
40
15
6
9
42
10
24
5
31
17
18
25
6
25
20
11
7
5
30
10
30
17
24
24
60
15
8
8
7
15
13
19
12
11
4
13
60
50
11
16
60
50
12
23
11
5
12
29
8
8
10
20
13
7
14
20
5
17
12
12
15
32
8
33
30
35
10
29
16
17
11
12
60
6
11
34
9
12
21
5
8
6
5
17
22
6
29
8
9
22
20
60
10
14
10
60
9
7
13
11
20
10
20
14
8
41
16
7
9
10
11
16
6
17
25
18
30
15
20
23
8
60
8
26
10
11
11
16
20
8
9
21
21
8
32
20
20
19
22
5
8
11
13
15
28
8
6
16
15
26
40
24
39
10
30
5
60
5
16
15
35
60
28
9
9
10
20
15
9
7
14
20
11
7
10
13
5
6
19
10
7
12
36
9
10
6
8
9
9
9
10
21
9
10
11
13
9
25
40
12
5
20
24
10
7
24
9
10
15
8
9
9
5
18
14
30
27
13
5
16
5
49
5
20
7
24
12
18
5
20
28
33
20
14
11
9
13
6
10
23
5
6
6
20
13
14
39
11
7
6
7
19
20
14
7
60
60
6
15
13
10
24
22
13
16
14
6
6
5
12
60
9
8
35
10
5
8
5
20
8
7
7
14
10
9
7
10
22
10
7
17
8
23
22
18
7
15
32
60
11
6
60
25
7
15
14
9
60
11
5
9
21
8
18
9
13
8
21
10
13
10
9
5
16
8
15
21
6
29
28
7
5
15
5
12
9
7
5
total task duration (303 Time Units [TU]), duration of the shortest task (5 TU), duration
of the longest task (60 TU) and purely linear flow: the output flow of operation u goes
directly into operation u 1, which induces a linear precedence graph. The
abovementioned models would naturally support other values for these invariants: this
is to test different configurations of task durations with a variable number of operators,
on linear or U-type layouts.
4.2 Experimentation results
Thereafter, only the result of case number 1 (Figure 5) will be detailed. For x operators,
the distribution of the 17 operations is given, for SALBP-2 (Lx) and SULBP-2 (Ux), as
well as the cycle durations for each operator and the takt time of the associated cell. Each
line also indicates the Chaku-Chaku of the operator concerned.
Variable takt
time
27
Figure 5.
Case number 1
results
The 34 studied cases are summarized in Table II. For each Lx and Ux configuration, the
cycle durations and the improvement percentage are given.
We observe that for one single operator, the results are the same in a straight or
U-shaped cell. This is a trivial result. This experiment is thus not significant.
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1 operator
U1
%
153
166
155
161
153
153
160
173
155
164
164
162
169
154
154
163
158
161
155
152
166
155
153
159
153
173
175
159
156
165
163
164
153
162
L2
152
152
152
152
152
152
152
152
152
152
153
152
155
152
154
152
153
152
153
152
152
152
153
153
153
153
152
153
153
152
152
153
152
153
0.65
8.43
1.94
5.59
0.65
0.65
5.00
12.14
1.94
7.32
6.71
6.17
8.28
1.30
0.00
6.75
3.16
5.59
1.29
0.00
8.43
1.94
0.00
3.77
0.00
11.56
13.14
3.77
1.92
7.88
6.75
6.71
0.65
5.56
2 operators
U2
%
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
303
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table II.
Summary of the
experimentation
results
L1
108
105
104
108
118
112
118
107
117
118
107
117
116
112
104
114
123
110
105
117
110
123
121
109
120
112
104
117
110
105
116
111
107
113
L3
103
102
102
102
103
103
104
102
104
104
103
103
105
103
104
104
103
102
103
103
104
103
102
102
102
102
104
102
102
103
102
103
102
103
4.63
2.86
1.92
5.56
12.71
8.04
11.86
4.67
11.11
11.86
3.74
11.97
9.48
8.04
0.00
8.77
16.26
7.27
1.90
11.97
5.45
16.26
15.70
6.42
15.00
8.93
0.00
12.82
7.27
1.90
12.07
7.21
4.67
8.85
3 operators
U3
%
80
96
85
83
78
95
86
88
87
84
82
84
86
81
96
85
85
91
90
84
85
90
81
85
90
87
89
83
81
88
84
86
83
85
L4
77
79
78
78
77
80
76
78
79
77
78
77
80
79
79
78
80
79
79
79
77
79
77
78
79
77
78
77
77
77
76
78
77
78
3.75
17.71
8.24
6.02
1.28
15.79
11.63
11.36
9.20
8.33
4.88
8.33
6.98
2.47
17.71
8.24
5.88
13.19
12.22
5.95
9.41
12.22
4.94
8.24
12.22
11.49
12.36
7.23
4.94
12.50
9.52
9.30
7.23
8.24
4 operators
U4
%
75
79
70
63
76
79
72
78
77
71
75
65
65
67
76
73
83
67
74
83
74
67
70
74
67
72
70
73
75
70
67
69
73
73
L5
64
62
65
62
63
62
62
64
63
64
63
63
63
63
63
62
63
62
62
62
65
63
61
65
63
63
62
62
62
63
64
66
62
64
14.67
21.52
7.14
1.59
17.11
21.52
13.89
17.95
18.18
9.86
16.00
3.08
3.08
5.97
17.11
15.07
24.10
7.46
16.22
25.30
12.16
5.97
12.86
12.16
5.97
12.50
11.43
15.07
17.33
10.00
4.48
4.35
15.07
12.33
5 operators
U5
%
65
61
70
60
60
60
62
63
63
60
60
60
63
60
60
60
63
60
60
60
67
63
61
65
60
60
68
62
62
63
64
68
61
68
L6
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
7.69
1.64
14.29
0.00
0.00
0.00
3.23
4.76
4.76
0.00
0.00
0.00
4.76
0.00
0.00
0.00
4.76
0.00
0.00
0.00
10.45
4.76
1.64
7.69
0.00
0.00
11.76
3.23
3.23
4.76
6.25
11.76
1.64
11.76
6 operators
U6
%
28
case no.
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
61
60
60
L7
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
60
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.64
0.00
0.00
7 operators
U7
%
IJLSS
6,1
For 2 or 3 operators, the results are equivalent in rare cases: the advantage generally is
in favor of U-Cells.
Similarly, for a higher number of operators, optimization is no longer efficient.
This can be easily explained, as the cycle duration is bounded by the bottleneck due to the
longest task assigned to the cell: Lx and Ux results converge, either for 7 operators, or
starting from 6 operators. These cases are not significant. Only Case 32 achieves a slight
performance advantage with 7 operators, advantage which is lost with 8 operators.
Between these two extreme cases (123 significant experiments), the results are
always in favor of a U-shaped layout, very often with improvements in cycle time
exceeding 10 per cent. Based on these significant experiments, 117 provide an
improvement of the performance (95 per cent). This result is substantial.
Variable takt
time
29
Table III.
Average and
standard deviation of
2 operators
3 operators
4 operators
5 operators
6 operators
L2
U2 % L3
U3 % L4
U4 % L5
U5 % L6 U6 % cycle durations of the
different
Average 160
153
5 112
103
8 86.1 78.0 9 72.4 63
13 62.7 60 7.7
configurations, for
SD
6.44
0.71
5.77
0.84
4.31 1.10
4.92 1.13
2.90 0
Lx and Ux (x [2, 6])
Figure 6.
Distribution of
results, for Lx & Ux
(x [2, 6])
IJLSS
6,1
30
i1
j1
i, j
j i
(17)
where fi,j 1 if a precedence relation exists between the two activities, and fi,j 0
otherwise.
From-to chart is the relationship flow matrix. It gives the dominance of the flows
between resources and indicates the closeness of resources in a facility layout design
(Mahdavi et al., 2008). From-to chart could be classified as loose and tight (Moore, 1980),
depending on whether it contains, respectively, a few and a large number of strong
precedence constraints. In our case, the From-to chart contains many entries with zero
values: this corresponds to the case of flow line layout (Hassan, 1994).
To minimize this moment, we proceed to activity permutations at the same level.
This allows us to find a local order which remains compatible with the previously
established global order.
The linear sequence which results from this treatment can be used as a reference
sequence to apply SALBP-2 and SULPB-2. Other sophisticated mathematical techniques
may also be applied, but what interests us here is the possibility to compare SALBP-2 and
SULBP-2 on equal and logical datasets. Thanks to this approach, we can consider that all
cases with complex precedence graphs can be transformed in cases with simple linear
precedence graphs. We can then use the approach mentioned in Sections 3 and 4.
Variable takt
time
31
Instance
Bowman8
Jackson
Jaeschke
Mansoor
Task
Task
No. of duration Average duration tasks
tasks
max
duration
min
durations
8
11
9
11
17
7
6
45
9.38
4.18
4.11
16.8
3
1
1
2
75
46
37
185
SD
Lin/
complex
Obtained sequence
4.31
1.94
1.45
14.8
C
C
C
C
1,2,4,3,6,5,8,7
1,2,3,4,5,7,6,8,9,10,11
1,2,3,4,5,6,7,8,9
3,1,2,4,5,6,7,8,9,10,11
Table IV.
Description of smallsized instances
IJLSS
6,1
Instance
Bowman8
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Jackson
32
Jaeschke
Table V.
Results for smallsized instances
Mansoor
75
75
0
46
46
0
37
37
0
185
185
0
42
38
10
24
24
0
20
19
5
98
93
5
28
26
7
19
13
32
14
13
7
68
65
4
22
21
5
13
12
8
10
10
0
50
49
2
20
18
10
12
10
17
17
17
0
11
9
18
9
8
11
7
7
0
45
45
0
For small-sized instances, the results confirm the performances achieved, similar to
those presented in Section 4. We will observe 32 per cent performance improvement for
the Jackson instance with three operators.
For large-sized instances, improvements are small for low-rate productions (number
of operators equal to two or three). On the other hand, for higher production rates (lower
takt time), improvements are substantial, easily exceeding 10 per cent. We also note that
the maximum production rate is more quickly reached with a U layout, as the maximum
rate limit, defined by the longest task duration, is reached with fewer operators: out of
the eight cases tested, we note that for Heskia, the highest rate is reached for a
U-shaped layout with 11 operators while 12 would be needed in a straight line; for
Gunther, Lutzl and Mitchell, the maximum rate is reached with 2 fewer operators;
and for Buxey, Kilbrid and Sauvey30, the maximum rate is reached with 3 fewer
operators, i.e. almost one-fifth of the payroll. This is a meaningful result.
7. Conclusion
All the experiments presented here show unequivocally that, with the same number of
operators, a cell in a U layout is always at least as efficient, in terms of reachable
production rates, than an equivalent cell in a linear layout. In summary, to produce more
with similar loads, U-Cell layouts are needed.
This was shown through our experiments and also with academic benchmarks.
Based on all tested cases (230 significant experiments), 96 per cent of cases give an
improvement of performance. This remains of course to be validated on industrial
situations.
The productivity gains observed are substantial. Furthermore, the performance
robustness obtained makes it possible to validate this approach in the case of a
manufacturing production subject to wide variations. Indeed, regardless of the rate
requested by the downstream production, the ad hoc configuration of the U layout will
be the most suitable for obtaining an optimal performance. This shows the great interest
of the U-shaped cell layout approach.
We conclude that U-Cells constitute a solution for implementing any relevant and
effective production system with variable structure. It is an essential tool for a lean
manufacturing approach.
Sauyer30
Roszieg
Mitchell
Lutz1
Kilbrid
Heskia
Gunther
Buxey
Instance
Task
duration
max
25
40
108
55
1400
13
13
25
No. of
tasks
29
35
28
45
32
21
25
30
10.8
441.9
12.27
36.57
13.8
11.17
Average
duration
100
Task
duration
(min)
324
125
105
14140
552
1024
483
324
tasks
durations
6.07
2.81
2.97
244.9
9.65
33.09
12.09
6.04
SD
Lin/complex
1,2,7,3,6,9,12,25,26,4,27,10,5,14,15,8,
13,19,11,16,21,17,18,22,20,23,24,28,29
17,1,2,5,10,12,3,6,4,7,8,9,11,14,18,13,
15,19,16,20,21,22,25,30,23,26,31,
24,32,27,33,28,34,29,35
2,1,3,4,5,6,8,17,19,21,22,23,24,26,7,9,
20,25,27,10,18,11,12,13,15,14,16,28
1,2,11,12,39,3,4,7,8,13,37,5,6,14,15,
43,9,10,16,17,18,23,24,25,29,30,31,32,
19,26,27,20,33,21,34,35,36,22,
28,38,40,41,42,44,45
3,4,2,1,5,6,9,8,7,10,11,12,13,15,14,16,
17,18,19,20,21,22,23,24,25,26,27,28,
29,30,31,32
1,2,3,4,21,5,6,7,8,14,9,10,11,12,13,15,
16,18,19,17,20
1,2,3,4,5,8,6,9,10,7,11,12,13,15,17,14,
16,19,20,23,18,21,22,24,25
1,10,3,2,4,5,11,12,16,17,6,13,18,7,14,
19,8,15,20,9,21,24,22,25,23,26,27,28,
29,30
Obtained sequence
Variable takt
time
33
Table VI.
Description of largesized datasets
Sauyer30
Roszieg
Mitchell
Lutz1
Kilbrid
Heskia
Gunther
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
%
Lx
Ux
Buxey
Table VII.
Results for largesized datasets
552
552
0
14140
14140
0
105
105
0
125
125
0
324
324
0
324
324
0
483
483
0
1024
1024
163
162
1
244
242
1
522
514
3
278
277
0
7296
7078
3
56
53
5
64
63
2
166
162
2
2
108
108
0
168
161
4
355
344
2
190
184
3
4920
4720
4
37
35
5
43
42
2
111
109
2
3
85
81
5
124
121
2
264
258
15
156
139
11
3702
3598
3
28
27
4
34
32
6
87
82
6
4
70
66
6
104
98
6
245
209
15
116
111
4
3100
2858
8
23
22
4
29
26
10
73
66
10
5
61
55
10
91
82
10
206
175
3
97
93
4
2654
2392
10
21
18
14
23
21
9
57
55
4
6
53
47
11
77
73
5
162
157
18
88
80
9
2284
2066
10
18
16
11
21
19
10
53
48
9
7
47
41
13
71
65
8
161
132
2
79
71
10
1928
1858
4
16
15
6
18
17
6
48
42
13
8
41
39
5
67
59
12
132
129
14
70
63
10
1898
1640
14
16
13
19
16
16
0
44
38
14
38
35
8
39
34
13
60
51
15
129
111
6
61
56
8
1676
1514
10
15
13
13
10
35
31
11
37
31
16
54
48
11
115
108
0
57
55
4
1628
1400
14
13
13
0
11
33
29
12
56
55
2
1430
1400
2
33
30
9
48
45
6
108
108
12
31
27
13
56
55
2
1400
1400
31
27
13
46
40
13
13
28
25
11
55
55
0
31
25
19
45
40
11
14
34
Instance
26
25
4
28
25
11
40
40
0
15
26
25
4
26
25
4
16
25
25
0
25
25
0
17
IJLSS
6,1
References
Aase, G.R., Olson, J.R. and Schniederjans, M.J. (2004), U-shaped assembly line layouts and their
impact on labour productivity: an experimental study, European Journal of Operational
Research, Vol. 156 No. 3, pp. 698-711.
Agpak, K. and Gken, H. (2007), A chance-constrained approach to stochastic line
balancing problem, European Journal of Operational Research, Vol. 180 No. 3,
pp. 1098-1115.
Alvarez, R., Calvo, R., Pea, M.M. and Domingo, R. (2009), Redesigning an assembly line
through lean manufacturing tools, International Journal of Advanced Manufacturing
Technology, Vol. 43 Nos 9/10, pp. 949-958.
Ariafar, S. and Ismail, D.N. (2009), An improved algorithm for layout design in
cellular manufacturing systems, Journal of Manufacturing Systems, Vol. 28 No. 4,
pp. 132-139.
Bartholdi, J. and Eisenstein, D. (1996), A production line that balances itself, Operations
Research, Vol. 44 No. 1, pp. 21-34.
Baybars, I. (1986), A survey of exact algorithms for the simple assembly line balancing,
Management Science, Vol. 32 No. 8, pp. 909-932.
Becker, C. and Scholl, A. (2006), A survey on problems and methods in generalized
assembly line balancing, European Journal of Operational Research, Vol. 168 No. 3,
pp. 694-715.
Boysen, N., Fliedner, M. and Scholl, A. (2007), A classification of assembly line
balancing problems, European Journal of Operational Research, Vol. 183 No. 2,
pp. 674-693.
Chiang, W. and Urban, T. (2006), The stochastic U-line balancing problem: a
heuristic procedure, European Journal of Operational Research, Vol. 175 No. 3,
pp. 1767-1781.
Crandall, K.C. (1973), Project planning with precedence lead/lag factors, Project Management
Quarterly, Vol. 4 No. 3, pp. 49-58.
Detty, R.B. and Yingling, J.C. (2000), Quantifying benefits of conversion to lean manufacturing
with discrete event simulation: a case study, International Journal of Production Research,
Vol. 38 No. 2, pp. 429-445.
Elmaghraby, S.E. (1977), Activity Networks: Project Planning and Control by Network Models,
John Wiley and Sons, New York, NY.
Esgin, E., Senkul, P. and Cimenbicer, C. (2010), A hybrid approach for process mining: using
from-to chart arranged by genetic algorithms, HAIS 2010, Part I, Lecture Notes on
Artificial Intelligence, Vol. 6076, pp. 178-186.
Guret, C., Prins, C. and Sevaux, M. (2002), Applications of Optimization with Xpress-MP, in
Translated and revised by Heipcke, S., Dash Optimization, Paris.
Hassan, M.M.D. (1994), Machine layout problem in modern manufacturing facilities,
International Journal of Production Research, Vol. 32 No. 11, pp. 2559-2584.
Hirst, P. and Zeitlin, J. (1991), Flexible specialization versus post-Fordism: theory, evidence and
policy implications, Economy and Society, Vol. 20 No. 1, pp. 5-9.
Holweg, M. (2007), The genealogy of lean production, Journal of Operation Management, Vol. 25
No. 2, pp. 420-437.
Irani, S. (1999), Handbook of Cellular Manufacturing Systems, Wiley, New York, NY.
Variable takt
time
35
IJLSS
6,1
36
Kara, Y., Paskoy, T. and Chang, C.T. (2009), Binary fuzzy goal programming approach to single
model straight and U-shaped assembly line balancing, European Journal of Operational
Research, Vol. 195 No. 2, pp. 335-347.
Kazemi, S., Ghodsi, R., Rabbani, M. and Tavakkoli-Moghaddam, R. (2011), A novel two-stage
genetic algorithm for a mixed-model U-line balancing problem with duplicated tasks,
International Journal of Advanced Manufacturing Technology, Vol. 55 Nos 9/12,
pp. 1111-1122.
Kim, Y.K., Kim, J.Y. and Kim, Y. (2006), An endosymbiotic evolutionary algorithm for the
integration of balancing and sequencing in mixed-model U-lines, European Journal of
Operational Research, Vol. 168 No. 3, pp. 838-852.
Kriengkorakot, N. and Pianthong, N. (2007), The U-line assembly line balancing problem, KKU
Engineering Journal, Vol. 34 No. 3, pp. 267-274.
Kulak, O., Durmusoglu, M.B. and Tufekci, S. (2005), A complete cellular manufacturing system
design methodology based on axiomatic design principles, Computers & Industrial
Engineering, Vol. 48 No. 4, pp. 765-787.
Kumar, M., Antony, J., Singh, R.K., Tiwari, M.K. and Perry, D. (2006), Implementing the lean
sigma framework in an Indian SME: a case study, Production Planning & Control, Vol. 17
No. 4, pp. 407-423.
Magee, D. (2007), How Toyota Became #1: Leadership Lessons from the Worlds Greatest Car
Company, Penguin Group, New York, NY.
Mahdavi, I., Shirazi, B. and Paydar, M.M. (2008), A flow matrix-based heuristic algorithm for cell
formation and layout design in cellular manufacturing system, International Journal of
Advanced Manufacturing Technology, Vol. 39 Nos 9/10, pp. 943-953.
Miltenburg, J. (1998), Balancing U-lines in a multiple U-line facility, European Journal of
Operational Research, Vol. 109 No. 1, pp. 1-23.
Miltenburg, J. (2001b). One-piece flow manufacturing on U-shaped production lines: a tutorial,
IIE Transactions, Vol. 33 No. 4, pp. 303-321.
Miltenburg, J. (2002), Balancing and scheduling mixed-model U-shaped production
lines, International Journal of Flexible Manufacturing Systems, Vol. 14 No. 2,
pp. 119-151.
Modarress, B., Ansari, A. and Lockwood, D.L. (2005), Kaizen costing for lean manufacturing:
a case study, International Journal of Production Research, Vol. 43 No. 9,
pp. 1751-1760.
Monden, Y. (1983), Toyota Production System, Industrial Engineering Press, Institute of Industrial
Engineers, Norcross.
Moore, J.M. (1980), The zone of compromise for evaluating lay-out arrangements, International
Journal of Production Research, Vol. 18 No. 1, pp. 1-10.
Motwani, J. (2003), A business process change framework for examining lean
manufacturing: a case study, Industrial Management Data Systems, Vol. 103 Nos
5/6, pp. 339-346.
Sabuncuoglu, I., Erel, E. and Alp, A. (2009), Ant colony optimization for the single model U-type
assembly line balancing problem, International Journal of Production Economics, Vol. 120
No. 2, pp. 287-300.
Scholl, A. and Klein, R. (1999a), Balancing assembly lines effectively - a computational
comparison, European Journal of Operational Research, Vol. 114 No. 1, pp. 50-58.
Scholl, A. and Klein, R. (1999b), ULINO: optimally balancing U-shaped JIT assembly lines,
International Journal of Production Research, Vol. 37 No. 1, pp. 721-736.
Shah, R. and Ward, P. (2003), Lean manufacturing: context, practice bundles, and performance,
Journal of Operations Management, Vol. 21 No. 2, pp. 129-149.
Singh, B., Garg, S.K., Sharma, S.K. and Grewal, C. (2010), Lean implementation and its
benefits to production industry, International Journal of Lean Six Sigma, Vol. 1 No. 2,
pp. 157-168.
Spengler, T., Volling, T. and Rehkopf, S. (2005), Zum Einsatz von Chaku-Chaku-Systemen in der
Montage konsumentennaher Erzeugnisse eine Fallstudie bei Rahmenauftragsfertigung,
Supply Chain Management und Logistik, Vol. 2 No. 1, pp. 249-275.
Sward, K. (1989), The Legend of Henry Ford, Easton Press, Norwalk.
Tavares, L.V. (1990), A multi-stage non-deterministic model for project scheduling under
resources constraints, European Journal of Operational Research, Vol. 49 No. 1,
pp. 92-101.
Wenming, Z. and Deuse, J. (2009), Cell staffing and standardized work design in Chaku-Chaku
production lines using a hybrid optimization algorithm, Proceeding of Computers &
Industrial Engineering, Troyes, pp. 305-310.
Weston, R.H. and Cui, Z. (2008), Next generation manufacturing systems, in Yan, X.T.,
Eynard, B. and Ion, W.J. (Eds), Global Design to Gain a Competitive Edge: An Holistic and
Collaborative Design Approach based on Computational Tools, Springer, London,
pp. 701-710.
Wiest, J.D. (1981), Precedence diagramming method: some unusual characteristics and their
implications for project managers, Journal of Operations Management, Vol. 1 No. 3,
pp. 213-222.
Yin, Y. and Yasuda, K. (2006), Similarity coefficient methods applied to the cell formation
problem: a taxonomy and review, International Journal of Production Economics, Vol. 101
No. 2, pp. 329-352.
Further reading
Chan, J.S., Samson, D.A. and Sohal, A.S. (1990), An integrative model of Japanese manufacturing
techniques, International Journal of Operations and Production Management, Vol. 10 No. 9,
pp. 37-56.
Liker, J. (2004), The Toyota Way, 14 Management Principles from the Worlds Greatest
Manufacturer, McGraw-Hill, New York, NY.
Miltenburg, J. (2001a). U-shaped production lines: a review of theory and practice, International
Journal of Production Economics, Vol. 70 No. 3, pp. 201-214.
About the authors
Patrick Pujo is an Associate Professor at Aix Marseille University of Marseilles (Centre de
Recherche sur le Transport et la LOGistique: CRET-LOG France) where he teaches Industrial
Engineering (Production Management, Quality Management, Production Systems Design, Line
Balancing, Lean Management and 6 []). He obtained a master degree then a PhD in Industrial
Engineering from the same university in 1994. His main research interests are Organization and
Decentralized Management of Production Systems, Simulation, Holonic Manufacturing System,
Isoarchic Control of Production System and Radio Frequency Identification (RFID) for Production
Tracking and Lean Manufacturing. He is the Associate Editor of the Journal europen des
systmes automatiss. Patrick Pujo is the corresponding author and can be contacted at:
patrick.pujo@univ-amu.fr
Variable takt
time
37
IJLSS
6,1
38
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com